From Historical Documents To Social Network Visualization: Potential Pitfalls and Network Modeling

Alexis Pister (1, 2) , Nicole Dufournaud (3) , Pascal Cristofoli (4) , Christophe Prieur (3) , Jean-Daniel Fekete (1)

1 AVIZ - Analysis and Visualization
2 IP Paris - Institut Polytechnique de Paris
3 Université Gustave Eiffel
4 EHESS - École des hautes études en sciences sociales

We describe the workflow followed by historians when conducting a Historical Social Network Analysis (HSNA) with five steps: textual sources acquisition, digitization, annotation, network creation, and analysis/visualization. While most analysis and visualization tools only support the last step, we argue that addressing the 2-3 last steps would boost the humanists' analytical capabilities. We explain why the network modeling process is particularly challenging and can lead to distortions of the sources, biases, and traceability problems. We list three main properties that we believe the constructed network should satisfy: alignment with reality/documents (not only with concepts), traceability (from documents to analysis/visualization and back), and simplicity (understandable by most and not more complex than needed). We claim that the model of bipartite dynamic multivariate network with roles allows an effective annotation/encoding of historical sources while satisfying these properties. We provide real-world examples of how this model has been used to answer socio-historical questions using visual analytics tools.

Mots clés: Applied computing; Arts and humanities; Human-centered computing; Visualization; Visualization application domains; Visual analytics

Domaines: Interface homme-machine [cs.HC]

Alexis Pister, Nicole Dufournaud, Pascal Cristofoli, Christophe Prieur, Jean-Daniel Fekete. From Historical Documents To Social Network Visualization: Potential Pitfalls and Network Modeling. VIS4DH 2022 - 7th Workshop on Visualization for the Digital Humanities, Oct 2022, Oklahoma City, United States. ⟨hal-03784532⟩